EF04 Topical Group Community Meeting

US/Eastern
Description

Connect through Zoom: https://cern.zoom.us/j/95541369778

Meeting ID: 955 4136 9778

Password: Ask coordinators

Michael Peskin - Global SMEFIT it Team
--------------------------------------
why SMEFT fits?
- test validity of SM: addition of precision measurements increases power of test

- connect deviations from SM to interpretations in terms of specific BSM models

- extract Higgs boson couplings: cannot measure Higgs width directly, need model to relate Higgs BR to Higgs partial widths and couplings; good even when butchering SMEFT (e.g., truncating at dim-6)

goals of SMEFT fit team

prepare illustrative global Higgs/EW fit including future lepton colliders and combinations of lepton+hadron colliders
want to understand better how to do SMEFT analysis, figure out subtleties and general issues of global fitting
illustrative = a concrete realization of fit, to be critiqued and improved as needed

current members: Jorge de Blas, Christophe Grosjean, Jiayin Gu, Michael Peskin, Junping Tian
all involved with previous SMEFT projections: ECFA Higgs report, ILC white papers (links)
links to papers describing approach used in various reports

different philosophies: EPS study tried to put together largest set of hadron and lepton collider projections; ILC study used only e+e-
also technical differences

one particular outcome of fits for which the two methods disagree: 1-loop effect of Higgs self-coupling
error on C6 in C6-only fit vs full SMEFT fit: different between ECFA and ILC, and different by factor 2 in C6 vs full SMEFT fit

another example: e+e- -> W+W-
full ILC simulations by Rosca, Marchesini, List
used dependence on 3 of the 5 decay angles
fit WWV vertices assuming SU(2)xU(1), three parameters: g(Z), k(A), lambda(A)
SMEFT predicts corrections to eeZ and enuW vertices that were not included; one could also increase sensitivity by using all 5 angles

two different methods to approach this:
ILC: use extrapolations to estimate result for 5 angles, define modified effective parameters (to include leading effects of electron vertex corrections)
ECFA: ignore full simulation, use optimal observable formalism to estimate best possible results

spent time discussing optimal observable formalism in team; it has power of giving lots of insight
this is goal of team (insight, not precise determinations)

optimal observables: Diehl, Nachtmann, Gunion, Grzadkowski
idealized version of unbinned likelihood analysis, useful for strictly linear SMEFT
here interested in fully differential cross section, to be integrated in full phase space (5D here)
expected error squared becomes ratio of integrals of cross section and variations on cross section; ultimately proportional to luminosity (i.e., error propto 1/sqrt(N))
best constraint becomes integral over phase space of product of variations divided by SM part of cross section

interesting formula: one can add dependent variables as wished
changes of variables and linear dependences can be implemented using linear algebra
constraints and efficiencies can be implemented by restricting the phase space (e.g., obvious cut on theta and leptons accounts for most of event reconstruction inefficiencies)

simple comparison to full sim, diagonal errors (in units of 1e-4)
much simpler than full sim, but has correct scaling with energy and number of phase space variables?
now interacting with Jenny List's group to compare results

use 10 variables for inclusion in SMEFT fit (e, gL, gR, gZ, gW, kA, kZ, lambdaA, lambdaZ, BR(W->lnu))
need also to add two nuisance parameters (dN, dPeff: luminosity and polarization): e+e- -> W+W- process used to determine effective luminosity and polarization: need to compute dependencies and marginalize

note: this is not replacement for full sim study, but simple tool usable to gain insight

turns out that replacing previous ILC inputs with this method does not change global Higgs coupling fit: still learning process

more questions
both analyses use strictly linear dependence on dim-6 SMEFT coefficients, and tree-level expressions for differential dependencies; should improve?
not obvious;
at this point, SMEFT fit closes
higher order perturbation theory brings in additional dim-6 operators
quadratic terms in dim-6 operators are at the same level as terms linear in dim-8 operators
another issue is that number of operators increases exponentially with dimension (and there are already a lot of dim-6 operators...)
adding these new operators without assumptions on their magnitude will make the fit not close

real conflict between trying to use best information and doing a calculation that closes
in past, introduced S and T formalism: inadequate to describe SM EW, but useful to deal with dim-6 operators

lepton vs hadron collider
lepton has advantages: no need to deal with QCD...
EW are at % level in one of the terms that appear in the formula about errors
Lambda is the small scale of new physics, can keep leading terms
situation different at LHC, philosophy "energy helps accuracy" does not help?

hope that for next-gen experiments, the simple, linear, lowest-order treatment of Delta_m (term in error formula) will be enough

lots of work to be done, team gotten started, hope to advance the art of SMEFT

Graham Wilson: 13, luminosity and polarization also constrained by other processes
MP: absolutely; they get input from other processes, want to have sort of global fit that check on those systematics

Ayres: comment about dim-8 operators; for Snowmass, not in business of analyzing data, but just doing projections (including HL-LHC); full global fit with (partial list of) dim-8 operators is daunting, 
what are examples of physics in which we can get qualitative contributions from dim-8? there could still be dim-6 couplings (to fermions); question: do we lose completely information about the type of 
operator we do not include (dim-6 fermion vs dim-8)?
MP: effects on shapes of distributions of multi-boson not captured by dim-6

Ilaria Brivio - LHC EFT WG
--------------------------
ATLAS, CMS, LHCb, Theory conveners

goals:
- facilitate interpretation of available measurements (from Higgs to top to EW)
- provide recommendations for the use of EFT by experiments, and platform for theoretical discussions
in practice:
ATLAS+CMS combined analysis of Higgs+EW+top measurements (so far done only Higgs coupling)
incorporation of LHCb data
help theory studies (translations, databases); coordinate theory studies

five working groups:
- EFT formalism
- predictions and tools
- experimental measurements
- fits and related systematics
- benchmark scenarios from UV models
- (heavy) flavor
  - part of flavor discussion is on formalism, part is on inclusion of b physics (LHCb)
link to google doc with outline of targets

- formalism:
recommendation for EW input parameters: mW, mZ, GF are current preferred set
other sets for scheme-dependency tests (e.g., alpha_EM, mZ, GF)

truncation uncertainties: recommend to report measurement as function of upper sliding cut Emax (clipping) vs. introduce theory uncertainty

recommendation for flavor assumption: most likely use U(2) for up quarks, and U(3) for down quarks (top is free)

need to discuss details of other symmetries, e.g., CP

- predictions and tools:
comparison and validation of prediction tools: SMEFTsim vs SMEFT@NLO
want to extend to other generators (SHERPA, POWHEG-BOX, JHUGen, VBF@NLO)
extend to published results, full simulation, on-shell + decay, NLO...

prediction database for published results w/ dynamical list of references with structured info (operators, processes, order, assumptions, scheme...)

NLO issues: k factors non-universality, running of Wilson coefficients, flavor treatment (asymmetries), uncertainties

- experimental measurements
survey of processes and operators, of observables (maximize sensitivity, define optimal observables for EFT)

determine relative sensitivities (e.g., Fisher information)
find measurement most constraining for a given operators

evaluate analysis techniques (inclusive vs fiducial vs differential, matrix element method, machine learning...)

- fits
feedback from ATLAS and CMS
comparison of fitting tools: EFTfitter, Fitmaker, HEPfit, SFitter, SMEFiT
use fixed benchmark scenarios, provide recommendations for common output format, try and use to incorporate LEP constraints

recommendation for output: covariance matrices, separate error sources, full likelihood?

long-term goal: define robust procedure for EFT combinations in ATLAS+CMS

fitting exercise:
global analysis project for the next 1 year
align assumptions in workspaces
align technical aspects
help convergence toward Higgs+EW+top ATLAS+CMS combination
favoring internal ATLAS+CMS combination at first, then do fit with public data and update with HL projections

- benchmark scenarios
define list of standard benchmark UV models
fully match at 1-loop to SMEFT
validate automated matching tools (SuperTracer, STrEAM, Matchmaker)

match MSSM in decoupling limit to SMEFT at 1-loop
advantage of MSSM: many full model results available; need to choose set of benchmark points

- flavor
discussing about activities
scope: how to include flavor constraints? explore LHCb input
platform for theory discussion on SMEFT vs WET
explore potential input from lower-energy / flavor experiments

Junping: practical question, scope is longer term than Snowmass, and effort aiming at being on Snowmass time line? E.g., predictions for HL-LHC, could they be done? benchmarks could also be useful
IB: benchmarks: time line reasonable; HL-LHC: less clear; fitting exercise: really need 1 year, bigger constraints are experiments
There are minutes attached to this event. Show them.